JPH11225120A - Communication system with improving method sound quality deteriorated in vox processing - Google Patents

Abstract

PROBLEM TO BE SOLVED: To hasten the judgement of sound/silence without increasing a delay by allowing a mobile station to advance an analyzing section to judge sound/ silence by several samples. SOLUTION: In processing voice decoding side including VOX processing, a sound/silence detecting part 410 analyzes received data to judge silence after receiving a postamble and to judge sound after receiving a preamble. When a front frame is judged to be sound and a present frame is judged to be sound at this time, a sound processing selecting part 411 selects a sound voice processing part 412 to output a sound decoding voice. On the other hand, when the front frame is judged to be sound but a present frame is judged to be silence, the part 411 selects a silence voice processing part 413 to execute sound interpolation processing to smooth decoded voice outputted at the time of shifting to silence from sound. In addition, when the front frame is judged to be silence but the present frame is judged to be sound, silence interpolation processing is executed.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a speech coding apparatus applied to various communications for transmitting and receiving information via communication circuits and various communication lines. Further, the present invention can be applied to a voice coding algorithm that codes every frame of a certain period and pre-reads data.

[0002]

2. Description of the Related Art VOs used in conventional communications
The X processing is shown below. (Digital automobile telephone system standard RCR STD-27E) The mobile station during communication turns on / off the transmission output of the user information transfer channel (hereinafter referred to as TCH) according to the presence or absence of voice.
F function. When detecting silence in a voiced state, the mobile station gives notice of the TCH transmission output OFF by a postamble (hereinafter referred to as PST). When a sound is detected in a silent state, a preamble signal (hereinafter referred to as PR)
E), the transmission output ON of the TCH is notified in advance.
The PST includes information on the mobile station background noise so that the base station can generate background noise during the silent period. The mobile station performs PST periodically (maximum once per second) during the silent period.
Is transmitted, and the background noise can be updated on the base station side. P
FIG. 1 shows transmission timings of RE and PST.

[0003]

As described above, the conventional V
The OX processing has a function of determining sound / non-sound for each frame of a fixed cycle and turning on / off a transmission output.
Therefore, when the sound rises from silence to sound, the PRE must be transmitted in the frame at the time when the sound is determined to be sound, so that the output of the one-frame sound is delayed, the beginning of the sound is cut off, and the audible sound quality is reduced. However, there was a problem that the steel was deteriorated.

According to the present invention, the determination of a sound can be hastened by moving forward the analysis section for which the determination of the sound / non-sound is made by several samples. As a result, the PRE can be transmitted one frame earlier at the rise from silence to a sound, and the sound quality can be improved in terms of audibility.

[0005]

SUMMARY OF THE INVENTION A mobile station equipped with the VOX processing sound quality deterioration improving method of the present invention is applied to a voice coding algorithm for pre-reading data in order to advance an analysis section for judging presence / absence by a few samples. It is possible. Look-ahead of data is conventionally required for windowing processing of linear prediction analysis,
It was not used for sound / silence determination. In the present invention, the pre-read data is also set as the analysis section for determining the presence / absence of sound / non-speech.
It is possible to hasten the determination of silence.

[0006]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Embodiments of the present invention will be described below with reference to the drawings, taking a cellular phone system as an example. FIG. 2 shows the configuration of the mobile phone system according to the present invention. The mobile phone system includes mobile terminals 200-2.
02, base stations 203 to 205, and an exchange control station 206. The mobile phone system is connected to a telephone 208 and other communication devices via a fixed telephone network 207. The VOX processing unit of the present invention is included in a mobile terminal, a base station, or an exchange control station as shown in FIG.

FIG. 3 shows an example of a specific configuration of the portable terminal shown in FIG. The portable terminal of the present invention includes a control unit 300 for performing overall control, a storage unit 301 for storing control data and control procedures, a keypad 302 for key input, and a display 30 for display output.
3, a sounder 304 for outputting a ring tone or the like, an interface unit 305 between these input / output units and the control unit, a wireless circuit unit 306, a baseband processing unit 307, a microphone 308, a speaker 309, It has an antenna 310. Each component is, for example, DSP, C
It can be realized by an electronic device such as a PU, a ROM, a RAM and various CMOSs.

FIGS. 4 (a) and 4 (b) show the V
1 shows an example of a speech encoding system including OX processing. FIG. 5 is a diagram showing a specific processing procedure of the VOX processing shown in FIG.

The processing on the voice encoding side including the VOX processing will be described below. First, the sound / non-sound detection unit 400 performs the sound / no-sound determination by calculating the power or the like of the input voice. The following data is transmitted according to the sound / silence determination result. If the previous frame is determined to be voiced and the current frame is determined to be voiced by the voiced / silence detection unit 400, the voice processing selection unit 401
02 is selected and the voice coded bit is transmitted.

If the previous frame is determined to be voiced and the current frame is determined to be silent by the voiced / silence detecting unit 400, the voice processing selecting unit 401
Select 3 and send postamble. When the previous frame is determined to be silent and the current frame is determined to be voiced by the voiced / voiceless detection unit 400, the voice processing selection unit 401 selects the voiced voice processing unit 402 and sends out a preamble.

If the previous frame is determined to be silent and the current frame is determined to be silent by the voiced / silent detection unit, and if a postamble has been transmitted to the previous frame, the voice processing selection unit 401 will determine that there is no sound. The voice processing unit 403 is selected, and a silent voice coded bit is transmitted.

When the previous frame is determined to be silent and the current frame is determined to be silent by the voiced / silence detecting unit 400, the postamble is not transmitted to the previous frame, and the postamble is transmitted. If more than one second has elapsed since then, the sound processing selection unit 401
03 is selected and the postamble is transmitted.

When the previous frame has been determined to be silent and the current frame has been determined to be silent by the voiced / silence detecting unit, and the postamble has not been transmitted to the previous frame and the postamble has been transmitted. If one second or more has not elapsed, the audio processing selection unit 401 selects the silent audio processing selection unit 403 and stops transmission.

The processing on the audio decoding side including the VOX processing will be described below. First, the sound / silence detection unit 410 performs sound / silence determination by analyzing the received data. The analysis method is silence after receiving the postamble,
After receiving the preamble, it is determined that there is sound. The data is decoded according to the following method based on the voiced / silent determination result.
If the previous frame is determined to be voiced and the current frame is determined to be voiced by the voiced / silent detection unit 411, the voice processing selection unit 411 selects the voiced voice processing unit 412,
Outputs voiced decoded speech.

If the previous frame is determined to be voiced and the current frame is determined to be silent by the voiced / silence detecting section 410, the voice processing selecting section 411 outputs the voiceless voice processing section 41.
No. 3 is selected, and the sound interpolation processing is performed. At this time, the sound interpolation processing is performed in the soundless sound processing unit 413 in order to smooth the decoded sound output when shifting from sound to silence.

If the previous frame is determined to be silent and the current frame is determined to be voiced by the voiced / silence detecting unit 410, the voice processing selecting unit 411 sends the voiced voice processing unit 41
2 is selected and silent interpolation processing is performed. At this time, the voiced speech processing unit 412 performs the silence interpolation processing in order to smooth the decoded voice output when shifting from voiceless to voiced.

If the previous frame is determined to be silent and the current frame is determined to be silent by the voiced / silent detection unit 410, and if a postamble has been received in the previous frame, the audio processing selection unit 411 Silent sound processing unit 413
Is selected, silent voice decoding is performed, and the decoded voice is output. When the previous frame is determined to be silent and the current frame is determined to be silent by the voiced / silent detection unit, and no postamble has been received in the previous frame, and at least one second has elapsed since the postamble was received. If the time has elapsed, the audio processing selection unit 411 selects the silent audio processing unit 413,
Outputs silence decoded voice.

When the previous frame has been determined to be silent and the current frame has been determined to be silent by the voiced / silent detection unit, and the postamble has not been received in the previous frame and the postamble has been received. If one second or more has not elapsed, the audio processing selection unit 411 selects the silent audio processing unit 413, performs a silent interpolation process, and outputs a decoded audio.

Next, the VOX processing method of the present invention is shown in FIG.
The VO of the present invention is shown in FIGS.
The X processing will be described. The mobile station during communication has a function of turning on / off the transmission output of a user information transfer channel (hereinafter referred to as TCH) according to the presence or absence of voice. When the mobile station detects silence in a voiced state, it transmits a TCH transmission output O by a postamble signal (hereinafter referred to as PST).
Announce FF. When a sound is detected in a silent state, a preamble signal (hereinafter, referred to as PRE) is used to detect TC.
The H transmission output ON is announced. The PST includes information on the mobile station background noise so that the base station can generate background noise during the silent period. The mobile station transmits the PST periodically (maximum once per second) during the silent period, so that the base station can update the background noise. FIG. 7 shows the transmission timing of the PRE and PST of the present invention.

The conventional VOX process shown in FIGS. 1A to 1C has a function of making a sound / no-sound determination for each frame of a fixed cycle and turning on / off the transmission output. Therefore,
When the sound rises from silence to sound, the PRE must be transmitted in the frame at the time when the sound is determined to be sound, so that the output of one-frame sound is delayed, the beginning of the sound is cut off, and the sound quality deteriorates in the sense of hearing. There was a problem to do.

In the present invention, the determination of the presence of a sound can be hastened by moving the analysis section in which the determination of the presence or absence of the sound is made forward by several samples. As a result, the PRE can be transmitted one frame earlier at the rise from silence to a sound, and the sound quality can be improved in terms of audibility. Moreover, when applied to a speech coding algorithm that prefetches data,
It is possible to hasten the determination of the presence / absence of sound without increasing the delay. Conventionally, data look-ahead has been required for linear prediction analysis processing, and has not been used for sound / silence determination. In the present invention, the pre-read data is also sound /
The analysis section was used to determine silence.

Next, an example of a speech coding algorithm for prefetching data will be described below. (Digital car telephone system standard RCR STD-27E) The half-rate speech coding algorithm of this system is based on pitch synchronous noise excitation source-code excitation source linear predictive coding (PSI-C).
ELP). PSI-CELP is a method of compressing the redundancy of speech by sending codebook parameters for generating speech, and further, by quantizing by processing the codebook in synchronization with the fundamental period of speech. It is characterized by improved efficiency.

FIG. 8 shows the overall configuration of the speech coding algorithm. For the encoder, 8 kHz sampling, 300
The input voice in the frequency band from Hz to 3.4 kHz is processed with one frame of 40 ms as one block. In each frame, parameters are obtained every 10 ms. First, quantization of LSP parameters and quantization of absolute power are performed by linear prediction analysis.

Next, quantization of the excitation code vector for generating speech is performed. The determination of these excitation sound sources is performed by calculating an auditory weighted process and optimizing the combination of the excitation coding vector and the gain. The obtained speech coded information is subjected to error correction, and a code for each frame is sent to the transmission path. After the error correction processing, the decoder decodes each parameter and synthesizes a waveform similarly to the encoder. Next, the synthesized waveform is passed through a post filter to obtain a final audio output.

The look-ahead of the data is L by linear prediction analysis.
This is performed at the time of SP quantization. The linear prediction analysis of the present speech coding algorithm will be described below. As shown in FIG. 9, a linear prediction analysis is performed on the input signal for each subframe by the autocorrelation method. Number of subframes is 4, subframe length is 8
0, the analysis section length is 160, the analysis center is the center of each subframe, and the prediction order is 10. In order to perform the linear prediction analysis of the third sub-frame, 40 sample data is pre-read from the frame processing.

Other voice coding algorithms for prefetching data include CS-ACELP and the like. These algorithms can be used for voice / audio without increasing delay.
The idea of the present invention that can hasten the determination of silence can be applied.

[0027]

According to the present invention, the determination can be hastened by moving the voiced / silent analysis section forward. As a result, the transition from sound to silence or from silence to sound can be accelerated, and sound quality degradation can be prevented.

Further, when the present invention is applied to a voice coding algorithm for prefetching data, it is possible to prevent sound quality deterioration without increasing delay.

[Brief description of the drawings]

FIG. 1 is a diagram showing transmission timings of conventional PRE and PST.

FIG. 2 is a configuration diagram of a mobile phone system according to an embodiment of the present invention.

FIG. 3 is a configuration diagram of a portable terminal of the present invention.

FIG. 4 is a configuration diagram of a VOX processing unit of the present invention.

FIG. 5 shows a specific processing procedure (COD) of the VOX processing of the present invention.
(ER side).

FIG. 6 shows a specific processing procedure (DEC) of the VOX processing of the present invention.
3 is a flowchart illustrating the operation of the ODER side.

FIG. 7 is a characteristic diagram showing an outline of VOX processing of the present invention.

FIG. 8 is a diagram showing an overall configuration of a PSI-CELP speech coding algorithm of the present invention.

FIG. 9 is an explanatory diagram of a linear prediction analysis according to the present invention.

[Explanation of symbols]

1-3: mobile terminal, 203-205 ... base station 2-
3, 206: switching control station, 207: fixed telephone network, 208: telephone.

Claims (3)

[Claims] An audio coding apparatus used in a communication system comprising a base station and a mobile station communicating with the base station determines whether or not there is speech from input speech. It is possible to reduce the power consumption without performing the encoding process, reduce the amount of data to be transmitted to the transmission path, and move the analysis section in which the sound / non-sound determination is performed forward by several samples. A communication system provided with a VOX processing sound quality deterioration improving method characterized by the above-mentioned. 2. A sound / silence determination method according to claim 1.
A communication system equipped with a VOX processing sound quality deterioration improvement method, characterized in that determination of presence / absence of a sound is made earlier than before, and word beginning truncation occurring at a rising portion of a speech from a silence to a speech is reduced. 3. A speech / silence discrimination without increasing delay time by applying the VOX processing according to claim 1 to a speech coding algorithm for prefetching data in order to perform windowing of linear prediction analysis. A communication system equipped with a VOX processing sound quality deterioration improvement method, wherein the method is faster than before, and reduces word start truncation occurring at a rising portion of speech from silence to speech.